Bayesian experimental design for identification of model propositions and conceptual model uncertainty reduction

摘要: Abstract The lack of hydrogeological data and knowledge often results in different propositions (or alternatives) to represent uncertain model components creates many candidate groundwater models using the same data. Uncertainty head prediction may become unnecessarily high. This study introduces an experimental design identify each component decrease uncertainty by reducing conceptual uncertainty. A discrimination criterion is developed based on posterior probability that directly uses evaluate importance. Bayesian averaging (BMA) used predict future observation aims find optimal number location observations sampling rounds such desired met. Hierarchical (HBMA) adopted assess if highly probable can be identified reduced design. implemented a Baton Rouge area, Louisiana. We new network existing USGS wells. sources create multiple are geological architecture, boundary condition, fault permeability architecture. All possible solutions enumerated multi-core supercomputer. Several found achieve 80%-identifiable 5 years six or more HBMA result shows proposition for once achieved. variances predictions significantly decreased probabilities unimportant propositions.